6 files changed, 88 insertions, 125 deletions

diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index 4f59a87b4..64de7e425 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -2,8 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <array>
 #include <cstddef>
+#include <vector>
 #include <glad/glad.h>
 
 #include "common/logging/log.h"
@@ -30,9 +32,27 @@ bool TestProgram(const GLchar* glsl) {
     return link_status == GL_TRUE;
 }
 
+std::vector<std::string_view> GetExtensions() {
+    GLint num_extensions;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &num_extensions);
+    std::vector<std::string_view> extensions;
+    extensions.reserve(num_extensions);
+    for (GLint index = 0; index < num_extensions; ++index) {
+        extensions.push_back(
+            reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, static_cast<GLuint>(index))));
+    }
+    return extensions;
+}
+
+bool HasExtension(const std::vector<std::string_view>& images, std::string_view extension) {
+    return std::find(images.begin(), images.end(), extension) != images.end();
+}
+
 } // Anonymous namespace
 
 Device::Device() {
+    const std::vector extensions = GetExtensions();
+
     uniform_buffer_alignment = GetInteger<std::size_t>(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT);
     shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
     max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
@@ -40,6 +60,7 @@ Device::Device() {
     has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group &&
                           GLAD_GL_NV_shader_thread_shuffle;
     has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
+    has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
     has_variable_aoffi = TestVariableAoffi();
     has_component_indexing_bug = TestComponentIndexingBug();
     has_precise_bug = TestPreciseBug();
@@ -55,6 +76,7 @@ Device::Device(std::nullptr_t) {
     max_varyings = 15;
     has_warp_intrinsics = true;
     has_vertex_viewport_layer = true;
+    has_image_load_formatted = true;
     has_variable_aoffi = true;
     has_component_indexing_bug = false;
     has_precise_bug = false;
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index ba6dcd3be..bb273c3d6 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -38,6 +38,10 @@ public:
         return has_vertex_viewport_layer;
     }
 
+    bool HasImageLoadFormatted() const {
+        return has_image_load_formatted;
+    }
+
     bool HasVariableAoffi() const {
         return has_variable_aoffi;
     }
@@ -61,6 +65,7 @@ private:
     u32 max_varyings{};
     bool has_warp_intrinsics{};
     bool has_vertex_viewport_layer{};
+    bool has_image_load_formatted{};
     bool has_variable_aoffi{};
     bool has_component_indexing_bug{};
     bool has_precise_bug{};
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 0dbc4c02f..42ca3b1bd 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -211,14 +211,14 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
     const auto primitive_mode{variant.primitive_mode};
     const auto texture_buffer_usage{variant.texture_buffer_usage};
 
-    std::string source = "#version 430 core\n"
-                         "#extension GL_ARB_separate_shader_objects : enable\n"
-                         "#extension GL_NV_gpu_shader5 : enable\n"
-                         "#extension GL_NV_shader_thread_group : enable\n"
-                         "#extension GL_NV_shader_thread_shuffle : enable\n";
-    if (entries.shader_viewport_layer_array) {
-        source += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
-    }
+    std::string source = R"(#version 430 core
+#extension GL_ARB_separate_shader_objects : enable
+#extension GL_ARB_shader_viewport_layer_array : enable
+#extension GL_EXT_shader_image_load_formatted : enable
+#extension GL_NV_gpu_shader5 : enable
+#extension GL_NV_shader_thread_group : enable
+#extension GL_NV_shader_thread_shuffle : enable
+)";
     if (program_type == ProgramType::Compute) {
         source += "#extension GL_ARB_compute_variable_group_size : require\n";
     }
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 74cb59bc1..e6b36a0f2 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -19,6 +19,7 @@
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
+#include "video_core/shader/node.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace OpenGL::GLShader {
@@ -398,8 +399,6 @@ public:
                                                        usage.is_read, usage.is_written);
         }
         entries.clip_distances = ir.GetClipDistances();
-        entries.shader_viewport_layer_array =
-            IsVertexShader(stage) && (ir.UsesLayer() || ir.UsesViewportIndex());
         entries.shader_length = ir.GetLength();
         return entries;
     }
@@ -725,36 +724,20 @@ private:
             const char* image_type = [&] {
                 switch (image.GetType()) {
                 case Tegra::Shader::ImageType::Texture1D:
-                    return "image1D";
+                    return "1D";
                 case Tegra::Shader::ImageType::TextureBuffer:
-                    return "imageBuffer";
+                    return "Buffer";
                 case Tegra::Shader::ImageType::Texture1DArray:
-                    return "image1DArray";
+                    return "1DArray";
                 case Tegra::Shader::ImageType::Texture2D:
-                    return "image2D";
+                    return "2D";
                 case Tegra::Shader::ImageType::Texture2DArray:
-                    return "image2DArray";
+                    return "2DArray";
                 case Tegra::Shader::ImageType::Texture3D:
-                    return "image3D";
+                    return "3D";
                 default:
                     UNREACHABLE();
-                    return "image1D";
-                }
-            }();
-
-            const auto [type_prefix, format] = [&]() -> std::pair<const char*, const char*> {
-                if (!image.IsSizeKnown()) {
-                    return {"", ""};
-                }
-                switch (image.GetSize()) {
-                case Tegra::Shader::ImageAtomicSize::U32:
-                    return {"u", "r32ui, "};
-                case Tegra::Shader::ImageAtomicSize::S32:
-                    return {"i", "r32i, "};
-                default:
-                    UNIMPLEMENTED_MSG("Unimplemented atomic size={}",
-                                      static_cast<u32>(image.GetSize()));
-                    return {"", ""};
+                    return "1D";
                 }
             }();
 
@@ -765,8 +748,12 @@ private:
                 qualifier += " writeonly";
             }
 
-            code.AddLine("layout (binding = IMAGE_BINDING_{}) {} uniform "
-                         "{} {};",
+            std::string format;
+            if (image.IsAtomic()) {
+                format = "r32ui, ";
+            }
+
+            code.AddLine("layout ({}binding = IMAGE_BINDING_{}) {} uniform uimage{} {};", format,
                          image.GetIndex(), qualifier, image_type, GetImage(image));
         }
         if (!images.empty()) {
@@ -1234,28 +1221,13 @@ private:
     }
 
     std::string BuildImageValues(Operation operation) {
+        constexpr std::array constructors{"uint", "uvec2", "uvec3", "uvec4"};
         const auto meta{std::get<MetaImage>(operation.GetMeta())};
-        const auto [constructors, type] = [&]() -> std::pair<std::array<const char*, 4>, Type> {
-            constexpr std::array float_constructors{"float", "vec2", "vec3", "vec4"};
-            if (!meta.image.IsSizeKnown()) {
-                return {float_constructors, Type::Float};
-            }
-            switch (meta.image.GetSize()) {
-            case Tegra::Shader::ImageAtomicSize::U32:
-                return {{"uint", "uvec2", "uvec3", "uvec4"}, Type::Uint};
-            case Tegra::Shader::ImageAtomicSize::S32:
-                return {{"int", "ivec2", "ivec3", "ivec4"}, Type::Uint};
-            default:
-                UNIMPLEMENTED_MSG("Unimplemented image size={}",
-                                  static_cast<u32>(meta.image.GetSize()));
-                return {float_constructors, Type::Float};
-            }
-        }();
 
         const std::size_t values_count{meta.values.size()};
         std::string expr = fmt::format("{}(", constructors.at(values_count - 1));
         for (std::size_t i = 0; i < values_count; ++i) {
-            expr += Visit(meta.values.at(i)).As(type);
+            expr += Visit(meta.values.at(i)).AsUint();
             if (i + 1 < values_count) {
                 expr += ", ";
             }
@@ -1264,29 +1236,6 @@ private:
         return expr;
     }
 
-    Expression AtomicImage(Operation operation, const char* opname) {
-        constexpr std::array constructors{"int(", "ivec2(", "ivec3(", "ivec4("};
-        const auto meta{std::get<MetaImage>(operation.GetMeta())};
-        ASSERT(meta.values.size() == 1);
-        ASSERT(meta.image.IsSizeKnown());
-
-        const auto type = [&]() {
-            switch (const auto size = meta.image.GetSize()) {
-            case Tegra::Shader::ImageAtomicSize::U32:
-                return Type::Uint;
-            case Tegra::Shader::ImageAtomicSize::S32:
-                return Type::Int;
-            default:
-                UNIMPLEMENTED_MSG("Unimplemented image size={}", static_cast<u32>(size));
-                return Type::Uint;
-            }
-        }();
-
-        return {fmt::format("{}({}, {}, {})", opname, GetImage(meta.image),
-                            BuildIntegerCoordinates(operation), Visit(meta.values[0]).As(type)),
-                type};
-    }
-
     Expression Assign(Operation operation) {
         const Node& dest = operation[0];
         const Node& src = operation[1];
@@ -1809,6 +1758,19 @@ private:
         return {tmp, Type::Float};
     }
 
+    Expression ImageLoad(Operation operation) {
+        if (!device.HasImageLoadFormatted()) {
+            LOG_ERROR(Render_OpenGL,
+                      "Device lacks GL_EXT_shader_image_load_formatted, stubbing image load");
+            return {"0", Type::Int};
+        }
+
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        return {fmt::format("imageLoad({}, {}){}", GetImage(meta.image),
+                            BuildIntegerCoordinates(operation), GetSwizzle(meta.element)),
+                Type::Uint};
+    }
+
     Expression ImageStore(Operation operation) {
         const auto meta{std::get<MetaImage>(operation.GetMeta())};
         code.AddLine("imageStore({}, {}, {});", GetImage(meta.image),
@@ -1816,31 +1778,14 @@ private:
         return {};
     }
 
-    Expression AtomicImageAdd(Operation operation) {
-        return AtomicImage(operation, "imageAtomicAdd");
-    }
-
-    Expression AtomicImageMin(Operation operation) {
-        return AtomicImage(operation, "imageAtomicMin");
-    }
-
-    Expression AtomicImageMax(Operation operation) {
-        return AtomicImage(operation, "imageAtomicMax");
-    }
-    Expression AtomicImageAnd(Operation operation) {
-        return AtomicImage(operation, "imageAtomicAnd");
-    }
-
-    Expression AtomicImageOr(Operation operation) {
-        return AtomicImage(operation, "imageAtomicOr");
-    }
-
-    Expression AtomicImageXor(Operation operation) {
-        return AtomicImage(operation, "imageAtomicXor");
-    }
+    template <const std::string_view& opname>
+    Expression AtomicImage(Operation operation) {
+        const auto meta{std::get<MetaImage>(operation.GetMeta())};
+        ASSERT(meta.values.size() == 1);
 
-    Expression AtomicImageExchange(Operation operation) {
-        return AtomicImage(operation, "imageAtomicExchange");
+        return {fmt::format("imageAtomic{}({}, {}, {})", opname, GetImage(meta.image),
+                            BuildIntegerCoordinates(operation), Visit(meta.values[0]).AsUint()),
+                Type::Uint};
     }
 
     Expression Branch(Operation operation) {
@@ -2035,6 +1980,12 @@ private:
         Func() = delete;
         ~Func() = delete;
 
+        static constexpr std::string_view Add = "Add";
+        static constexpr std::string_view And = "And";
+        static constexpr std::string_view Or = "Or";
+        static constexpr std::string_view Xor = "Xor";
+        static constexpr std::string_view Exchange = "Exchange";
+
         static constexpr std::string_view ShuffleIndexed = "shuffleNV";
         static constexpr std::string_view ShuffleUp = "shuffleUpNV";
         static constexpr std::string_view ShuffleDown = "shuffleDownNV";
@@ -2172,14 +2123,14 @@ private:
         &GLSLDecompiler::TextureQueryLod,
         &GLSLDecompiler::TexelFetch,
 
+        &GLSLDecompiler::ImageLoad,
         &GLSLDecompiler::ImageStore,
-        &GLSLDecompiler::AtomicImageAdd,
-        &GLSLDecompiler::AtomicImageMin,
-        &GLSLDecompiler::AtomicImageMax,
-        &GLSLDecompiler::AtomicImageAnd,
-        &GLSLDecompiler::AtomicImageOr,
-        &GLSLDecompiler::AtomicImageXor,
-        &GLSLDecompiler::AtomicImageExchange,
+
+        &GLSLDecompiler::AtomicImage<Func::Add>,
+        &GLSLDecompiler::AtomicImage<Func::And>,
+        &GLSLDecompiler::AtomicImage<Func::Or>,
+        &GLSLDecompiler::AtomicImage<Func::Xor>,
+        &GLSLDecompiler::AtomicImage<Func::Exchange>,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.h b/src/video_core/renderer_opengl/gl_shader_decompiler.h
index 2ea02f5bf..e538dc001 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.h
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.h
@@ -90,7 +90,6 @@ struct ShaderEntries {
     std::vector<ImageEntry> images;
     std::vector<GlobalMemoryEntry> global_memory_entries;
     std::array<bool, Maxwell::NumClipDistances> clip_distances{};
-    bool shader_viewport_layer_array{};
     std::size_t shader_length{};
 };
 
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index f141c4e3b..6a7012b54 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -343,20 +343,17 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         u8 is_bindless{};
         u8 is_written{};
         u8 is_read{};
-        u8 is_size_known{};
-        u32 size{};
+        u8 is_atomic{};
         if (!LoadObjectFromPrecompiled(offset) || !LoadObjectFromPrecompiled(index) ||
             !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_bindless) ||
             !LoadObjectFromPrecompiled(is_written) || !LoadObjectFromPrecompiled(is_read) ||
-            !LoadObjectFromPrecompiled(is_size_known) || !LoadObjectFromPrecompiled(size)) {
+            !LoadObjectFromPrecompiled(is_atomic)) {
             return {};
         }
         entry.entries.images.emplace_back(
             static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
             static_cast<Tegra::Shader::ImageType>(type), is_bindless != 0, is_written != 0,
-            is_read != 0,
-            is_size_known ? std::make_optional(static_cast<Tegra::Shader::ImageAtomicSize>(size))
-                          : std::nullopt);
+            is_read != 0, is_atomic != 0);
     }
 
     u32 global_memory_count{};
@@ -382,12 +379,6 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         }
     }
 
-    bool shader_viewport_layer_array{};
-    if (!LoadObjectFromPrecompiled(shader_viewport_layer_array)) {
-        return {};
-    }
-    entry.entries.shader_viewport_layer_array = shader_viewport_layer_array;
-
     u64 shader_length{};
     if (!LoadObjectFromPrecompiled(shader_length)) {
         return {};
@@ -435,14 +426,13 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         return false;
     }
     for (const auto& image : entries.images) {
-        const u32 size = image.IsSizeKnown() ? static_cast<u32>(image.GetSize()) : 0U;
         if (!SaveObjectToPrecompiled(static_cast<u64>(image.GetOffset())) ||
             !SaveObjectToPrecompiled(static_cast<u64>(image.GetIndex())) ||
             !SaveObjectToPrecompiled(static_cast<u32>(image.GetType())) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsBindless() ? 1 : 0)) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsWritten() ? 1 : 0)) ||
             !SaveObjectToPrecompiled(static_cast<u8>(image.IsRead() ? 1 : 0)) ||
-            !SaveObjectToPrecompiled(image.IsSizeKnown()) || !SaveObjectToPrecompiled(size)) {
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsAtomic() ? 1 : 0))) {
             return false;
         }
     }
@@ -464,10 +454,6 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         }
     }
 
-    if (!SaveObjectToPrecompiled(entries.shader_viewport_layer_array)) {
-        return false;
-    }
-
     if (!SaveObjectToPrecompiled(static_cast<u64>(entries.shader_length))) {
         return false;
     }